Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
Currently, when a trigger run has failed pipeline runs, users must manually identify and re-trigger each failed run individually. There is no built-in way to rerun a batch of failed pipeline runs from a specific point in the DAG.
Additionally, triggers are pinned to a specific pipeline revision. When a new revision is registered (e.g., with bug fixes or model improvements), users must manually update and recreate their trigger runs to pick up the new revision. The auto_flip field exists in the YAML schema and UI, but the runtime logic that automatically switches to the latest revision is not yet active.
Describe the solution you'd like
A clear and concise description of what you want to happen.
Batch Rerun — Add the ability to rerun a set of failed pipeline runs in bulk from a trigger run. Users should be able to resume from a specific point in the pipeline DAG rather than re-executing the entire pipeline, reducing wasted compute for transient failures. This could be exposed as a CLI command (e.g., ma trigger_run rerun --namespace= --name=) and through the UI.
Auto Flip — Activate the auto_flip runtime logic so that when spec.auto_flip: true is set on a TriggerRun, the trigger automatically picks up the latest pipeline revision whenever a new one is registered. This is critical for production pipelines that should stay current with the latest code, bug fixes, and platform library updates without manual intervention. The logic should:
Detect new pipeline revisions on the master/main branch
Automatically update the trigger's revision reference
Disable auto flip if the pipeline is not on the master branch (with a user-facing warning)
Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.
Manual rerun: Users manually create new trigger runs for each failed pipeline run — tedious and error-prone at scale.
Manual revision updates: Users manually kill old trigger runs and create new ones, whenever a new pipeline revision is registered — adds operational overhead and risks running stale code in production.
CI/CD-driven trigger updates: Use external CI/CD to detect new revisions and recreate triggers — adds infrastructure complexity outside the platform.
Additional context
Add any other context or screenshots about the feature request here.
Both features are already documented as "Coming Soon" in the Set Up Triggers guide and Automatic Revision Tracking (preview). The auto_flip field is already part of the TriggerRun protobuf spec and accepted in YAML.
Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
Currently, when a trigger run has failed pipeline runs, users must manually identify and re-trigger each failed run individually. There is no built-in way to rerun a batch of failed pipeline runs from a specific point in the DAG.
Additionally, triggers are pinned to a specific pipeline revision. When a new revision is registered (e.g., with bug fixes or model improvements), users must manually update and recreate their trigger runs to pick up the new revision. The auto_flip field exists in the YAML schema and UI, but the runtime logic that automatically switches to the latest revision is not yet active.
Describe the solution you'd like
A clear and concise description of what you want to happen.
Batch Rerun — Add the ability to rerun a set of failed pipeline runs in bulk from a trigger run. Users should be able to resume from a specific point in the pipeline DAG rather than re-executing the entire pipeline, reducing wasted compute for transient failures. This could be exposed as a CLI command (e.g., ma trigger_run rerun --namespace= --name=) and through the UI.
Auto Flip — Activate the auto_flip runtime logic so that when spec.auto_flip: true is set on a TriggerRun, the trigger automatically picks up the latest pipeline revision whenever a new one is registered. This is critical for production pipelines that should stay current with the latest code, bug fixes, and platform library updates without manual intervention. The logic should:
Detect new pipeline revisions on the master/main branch
Automatically update the trigger's revision reference
Disable auto flip if the pipeline is not on the master branch (with a user-facing warning)
Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.
Manual rerun: Users manually create new trigger runs for each failed pipeline run — tedious and error-prone at scale.
Manual revision updates: Users manually kill old trigger runs and create new ones, whenever a new pipeline revision is registered — adds operational overhead and risks running stale code in production.
CI/CD-driven trigger updates: Use external CI/CD to detect new revisions and recreate triggers — adds infrastructure complexity outside the platform.
Additional context
Add any other context or screenshots about the feature request here.
Both features are already documented as "Coming Soon" in the Set Up Triggers guide and Automatic Revision Tracking (preview). The auto_flip field is already part of the TriggerRun protobuf spec and accepted in YAML.